It is always important for clinicians to understand how patients are doing, not just those in crisis, but also those they aren’t alarmed about at any particular moment. But for pediatric patients, it’s even more so: first, because of the inherent sadness of unnecessary harm to a child, but second, because that harm may come on with little obvious warning.
Due to children’s physiology, the nature of physical improvement or deterioration is different than in adults. Children’s systems tend to compensate for a period of time but then rapidly decline when compensation is pushed beyond its limit. Furthermore, once children develop a serious problem, they often don’t have the physiological reserves to buffer its effects and recover.
After we (StevenRothman and I) developed the Rothman Index (RI), for adults, we approached the problem of applying the same methodology to pediatrics. Other generally available early warning models such as PEWS, suffer from multiple shortcomings: they are based primarily on vital signs, which tend to signal only when the patient is close to decompensation, and they employ artificial age brackets, creating discontinuities in the score. A useful model would take into account the patient’s functional status, as captured through nursing notes, and would provide a continuous adjustment of risk as a function of age.. In developing the pRI, we addressed these issues. We developed the pRI using data from 80,000 pediatric patient visits and collaborating with clinical experts from three hospital research partners: Children’s Hospital of Pittsburgh, Yale New Haven Children’s Hospital, and University of Florida Health. A Journal of Biomedical Informatics publication details the pRI development and validation work.
For all mammals – humans and animals alike – heart rate is a function of heat transfer, body mass and surface area, as the body tries to maintain a constant temperature. And so in children, with a high surface area to mass ratio, average heart rate is higher, and it also has more variability, than we find in adults. Given nuances like this, we examined all of the RI’s continuous variables for age dependence, and determined there are five variables with significant age dependencies: heart rate, respiration rate, systolic blood pressure, diastolic blood pressure, and serum creatinine. The first four have age-based dependencies relating to the basic physics described above. The last, creatinine, increases with age as muscle mass increases.
Our resulting pRI, now used by hospitals across the country, evaluates risk on a continual age-adjusted basis. It is sophisticated enough to adjust to changes in the young child’s body, like the significant weight gain from 0 to 5 years old, and yet accounts for older teenagers who more closely resemble adults. Although we use the same head-to-toe nursing assessment categories for children as we do for adults, the assessments are tailored, by the hospital, for age appropriate criteria.
Our recent FDA review process, resulting in FDA 510(k) clearance, included not only examination of the adult RI, but also a rigorous examination of the pRI’s performance across the pediatric patient spectrum. Patient acuity systems are standardly evaluated based on how well they predict death within 24 hours; it’s a performance metric, and similar to that for automobiles, time for acceleration from 0 to 60, it’s also not an indication of intended use. . Fortunately, mortality rates at pediatric hospitals are low. However good performance in this metric is indicative of a model with good general performance. Our reported results for predicting 24-hour mortality, as published, are excellent. For the FDA studies, we went further and broke mortality prediction analysis down by age group and over longer timescales (within 24, 48, and 72 hours) as well as separately for surgical and medical patients.
The analysis that we provided to FDA demonstrated: Excellent performance of the pRI in predicting mortality within 24, 48, and 72-hours across all age groups; the performance of the pRI for predicting unplanned transfer to ICU within 24 hours is strong across age groups; and, the performance of the pRI for predicting mortality remains excellent when assessed separately for medical and surgical sub-populations.
There is a growing literature of the use of our work for measuring patient condition in a wide-range of disease categories and care settings, and for its use toward our main goal, improving care. However, much remains to be learned. We really have only begun to understand the implications of this general measurement of patient condition, and we will keep striving to improve predictive capabilities and will maintain a robust research agenda. The Rothman Society for Innovation and Research (RSIR) partners with anyone wishing to engage in rigorous scientific discussion around refining and applying the RI.
One of our current research priorities is predicting and reducing unplanned readmissions. As we would expect, children differ from adults in drivers of readmission. I invite everyone to join us and register for a live webinar titled, “Predicting Unplanned Readmissions with the Pediatric Rothman Index,” on July 19, 2-3 p.m. ET. Nationally-recognized clinical informaticist and research partner Dr. William Feaster, Chief Medical Informatics Officer at Children’s Hospital of Orange County (CHOC), will join us with co-speaker Louis Ehwerhemuepha, PhD, a lead data scientist who helped develop CHOC’s model. Together, they will share the story behind model development, the data components that make up the calculation – including social determinants of health – and how, every morning, the model runs and provides a risk of readmission score to the care management team, which then implements agreed-upon interventions.
Congratulations to the research team, as the journal Hospital Pediatrics recently accepted the model for publication (date TBD), with the reviewers noting CHOC’s predictive tool is, “A novel model for enhanced prediction and understanding of unplanned 30-day pediatric readmission.” On the webinar, Drs. Feaster and Ehwerhemuepha will also review criteria for additional sites to participate in their next phase of research.
It is my honor to continue this work, and I look forward to sharing enhancements to the pRI as we learn more and more about helping save children’s lives.